Interesting. Usually, I would have expected Farkas to wiggle out with the help of $$ and serve little to no time, which is the case with almost every rich guy. But in this case, it seems that he is still serving time! From Wikipedia:

Seriously, what's the difference between lobbying and bribery?

Religion? Eating meat? Utter nonsense. I know this is Slashdot and no one RTFA, but if you have a claim, you can quickly look it up using the Ctrl-F supercombo key.

India is following a global trend to phase out animal dissection, or to make it voluntary, despite opposition from scientists who say that the experience is impossible to replicate with models.

India’s ban comes after pressure from animal-rights groups such as the Indian arm of PETA (People for the Ethical Treatment of Animals), based in Mumbai, which used Bollywood celebrities in a high-profile campaign to bring college students on side

Why do you think R is not easy to implement? My company has been using SAS for a long time and we are finally making the change to R. As far as OP's requirements are concerned, I think R is way superior to SAS or SPSS because of its free, modular nature. It is clean, simple and suitable for a wide range of users. The commercial packages are filled with way too much business lingo garbage for me.

I personally think commercial support is overrated. I can install software on my own. I know how to browse through manuals and other information to find what I need. For a package like R, I almost always get any questions answered in at most few hours on online forums. So what exactly do I get from commercial support for my money? But, if OP needs commercial support, there is an enterprise edition of R by Revolution Analytics located here: http://www.revolutionanalytics.com/products/revolution-enterprise.php. Might be worth looking into.

Bottom line: R all the way.

OP hasn't mentioned a lot except budget. Since you are on such a tight budget, I would highly recommend doing some theoretical analysis first. Do you have a serial code? How much parallelism exists in the code? You say the task is 'very parallel', but Amdahl's law (which is really common sense) will tell you that even for small amounts of serial sections of code, your speedup will be limited. You should also consider the amount of time the code actually runs. Achieving a speedup of 2 for a serial code that runs for one minute is near worthless.

After you estimate speedup, do some rough calculations on the basis of average cost of a processor and the the number of processors required. This should give you an estimate of the hardware cost required. Compare that with the cost of CPU cycles per dollar you get using a cloud service such as Amazon.

Like what?

Kevin Mitnick was recently on Colbert Report to promote his book. Here is the link if anyone's interested.

If Time Zones in Indiana are not broken, then I don't know what is. I mean, if the times could change if you commute from one place to another. Ridiculous.

Funny, as someone who is about to start working as a Statistician, all three courses add concrete value to my job - especially Machine Learning and Databases. Looking forward to the courses.

Yup, the physical distances cannot be ignored anymore. If fact, they try to place the servers as physically close to the corresponding stock exchange as possible in order to avoid speed-of-light delays!

This is purely my guesswork, but I think that the benefits would be marginal. An improvement of 1 fps might not be convincing for game devs to modify the kernel. However, if you can execute even a single trade before others do, and multiply it by the number of trades executed per second (which must be in millions), the profit margin would be non-trivial.

As mentioned in the summary, linux allows the firms to modify the OS kernel to serve their purpose. For example, the performance of trading algorithms would considerably degrade if context switching is allowed. So you can modify the kernel so as to dedicate certain cores for the main algorithms to which the OS can pass a very limited number of signals.

I know for a fact that at least one bank employs this in their high frequency trading group and probably all of them do.

To me, it seems like a simple Machine Learning problem. The number of common friends will obviously be the most influencing input parameter followed by school, work place, etc.

Unless someone makes me understand the complexities underneath, I am going to assume that anyone can solve this problem by taking one course in Machine Learning.

Link to the S&P report which contains their rationale for downgrading, future outlook, etc.